lag-Metal Equilibrium During Submerged

re Welding
=,.$

%
%

C. S. CHAI AND T. W. EAGAR

A thermodynamic model of the equilibria existing between the slag and the weld metal
during submerged arc welding is presented. As formulated, the model applies only to fused
neutral fluxes containing less than 20 pet CaF,, however some results indicate that the model
may be useful in more general cases as well. The model is shown to be capable of predicting
the gain or loss of both Mn and Si over a wide range of baseplate, electrode and flux
compositions. At large deviations from the predicted equilibrium, the experimental results
indicate considerable variability in the amount of Mn or Si transferred between the slag and
metal phases, while closer to the calculated equilibrium, the extent of metal transfer becomes
more predictable, The variability in metal transfer rate at large deviations from equilibrium
may be explained by variations between the bulk and the surface concentrations of Mn and
Si in both metal and slag phases.

SINCE

study of equilibrium thermodynamics, which is capable

its inception over forty years ago, submerged
of defining the direction and bounds of a reaction if not
arc welding has developed into one of the most
the actual extent of the reaction. Attempts have been
efficient, most reliable techiques of joining steel. Immade to utilize the slag-metal equilibrium of steelprovements in the flux and electrode compositions have
making with respect to the welding process.15 Unresulted in weld metal with excellent properties. Nonefortunately, these attempts have met with little success.
theless, the gains have been slow and have been won
In the present paper a new formalism is developed
with considerable difficulty which is in part due to the
from which to predict slag-metal equilibria during arc
fact that no formalism has been developed to explain
welding. The predicted equilibria are then compred with
the observed trends. Indeed, flux development for
experimental data taken both from welding literature
submerged arc and other welding processes has been
and from new experimental measurements. It will be
approached empirically leaving a limited foundation
shown subsequently that the new formalism is quite
upon which to base future progress.
successful in predicting the direction and the perA major advance in systemitizing the study of
missable extent of both the manganese and the silicon
welding fluxes was begun by Christensen and Chipman1
reactions. The actual extent of the reactions is deterwho investigated acidic coatings on shielded metal arc
mined by kinetic factors rather than by thermodynamic
electrodes. Others f o l l o ~ i n g , have
~ - ~ made similar studfactors and hence is more difficult to predict. Noneies on submerged arc welding fluxes, with some success
theless the present approach is useful in establishing
in categorizing the effects of the various flux compolimits upon the chemistry changes which occur during
nents. Several reviews have been written upon the
the submerged arc welding process.
subject?JOMore recently, several theories have been
proposed to explain the observed t r e n d ~ , ~ - howI~-~~-~~
ever, the conclusions are often conflicting, which leaves
the field at little advantage.
THEORETICAL APROACH
An important concern of many of the previous
studies has been quantification of the slag-metal reac1. Initial Assumptions
tions which occur during the welding process; however,
As noted previously, it is a generally accepted fact
the lack of a uniform reaction temperature and the brief
that the slag and metal do not achieve equilibrium
reaction time results in a nonequilibrium precess. The
failure to achieve equilibrium presents a major difficulty
during flux shielded welding. For this reason most
in establishing a formalism for predicting the extent of
investigators have chosen an empirical approach to
welding flux formulation. An important initial hypoththese reactions. Similar difficulties occur in other slagmetal processes such as steelmaking and smelting.
esis in the present work is the belief that equilibrium
thermodynamics, while unable to predict the extent of a
Much advantage has been gained in these areas through
reaction, should at least be able to predict the direction
of the reaction. It is further assumed that it is possible to
utilize much of the slag-metal activity data developed
C. S. CHAI, presently Engineer, Lincoln Electric Co.,Cleveland,
for steelmaking, realizing that the rate controlling
OH, was formerly a Graduate Research Assistant, Department of
Materials Science and Engineering, Massachusetts Institute of Techreactions may be different in welding. Using these
nology, Cambridge, MA 02139. T. W. EAGAR is Associate Professor
hypotheses, a model is developed which uses the
of Materials Engineering, Department of Materials Science and
following assumptions.
Engineering, Massachusetts Institute of Technology, Cambridge, MA
1) The effective temperature of chemical reactions in
02139.
Manuscript submitted July 16, 1980
the weld metal pool is 2000 OC.
ISSN 0360-2141 181 1091 1-0539S00.7510
METALLURGICAL TRANSACTIONS B @ 1981 AMERICAN SOCIETY FOR METALS AND VOLUME 12B. SEPTEMBER 1981-539
THE METALLURGICAL SOCIETY OF AIME

2) An empirical relationship between the basicity

index of the welding flux and the weld metal oxygen
content is known to exist and is valid for all flux
compositions of interest.
3) The activity data of steelmaking slag may be
extrapolated from 1600 OC to the 2000 OC range by
assuming regular solution behavior, i.e., T In a, equals a
constant.
4) The primary reactions of interest are those involving silicon, manganese and oxygen.
Each of these assumptions requires some justification.
Assumption 1. In his review of welding fluxes,
Jackson9 noted that most investigators have measured
the maximum weld pool temperature as 2000 OC using
thermocouple techniques. This temperature is an average value representing the median between the fusion
temperature (approximately 1500 OC) and the weld pool
surface temperature in the arc vicinity, which may be
roughly estimated from metal evaporation data as

BI =

CaF,

*This estimate is obtained by equating the heat lost by metal

evaporation as calculated by Cobine and BurgerMto the heat
transferred to the weld pool surface as measured by Nestor."

Chipman1 and Belton et a13 found an effective equilibrium temperature for the MnO and the SiO2reactions
respectively in selected fluxes, to be 2000 OC based
upon chemical analysis of the weld metal.
Perhaps more important than either of these justifications is the finding which will be presented subsequently that equilibria based upon 2000 OC agree with
the experimental data while equilibria based upon other
temperatures do not.
Assumption 2. Kubli and Sharav2 showed that the
oxygen content of submerged arc weld metal decreases
with increasing flux basicity. Tuliani et al,I8quantified
the oxygen content of the weld metal as a function of
the basicity index. The International Institute of Welding flux basicity index (BI) used by Tuliani et a1 is given
by:

+ CaO + MgO + BaO + SrO + N a 2 0 + K,O + Li20 + 1/2(MnO + FeO)

SiO, + 1/2(A1& + Ti02 + Zr02)

Eagari3has modified this relationship slightly by

omitting the CaF2 term. The relationship between the
weld metal oxygen content and the welding flux
basicity, based upon Tuliani's data and Eagar's BI is
shown in Fig. 1. Much more data could be added to this
graph, however, the trend would not be changed. The
oxygen content of welds produced with acid fluxes has
been found experimentally to be strongly dependent
upon the BI of the flux while the oxygen content of
basic fluxes is essentially independent of the BI. An
explanation for this trend has been given by Eagar.I3J4
In the model presented here, the oxygen content of a
given flux is estimated a priori from the solid line of
Fig. 1.
Asszimption 3. No experimental justification of this
assumption is available. This hypothesis is made as a
matter of convenience due to lack of any better
alternative. However it should be noted that the slag
acitivites are changed relatively little by this assumption. For example, the activity of SiO, in a 40 pct

2500 OC.* It should also be noted that Christensen and

0L

I
2
Flux B a s i c i t y Index

I
3

Fig. 1-The empirical relationship between the weld metal oxygen

and the flux basicity index when using the submerged arc welding
process. (After Tuliani, el a1 la)
540-VOLUME 12B. SEPTEMBER 1981

CaO-60 pet SiO-,melt is changed from 0.80 at 1600 OC

to 0.75 at 2000 OC by assuming regular solution behavior.
Assumption 4. The oxidizing potential of welding
fluxes is not controlled by the FeO content of the slag as
it is in ironmaking and steelmaking. Although fluxes
high in FeO produce large weld metal oxygen contents,'
the oxygen content of welds made with fluxes containing less than ten percent FeO is not noticeably
influenced by the FeO content of the f l ~ x . Eagar
~ J ~ has
shown that this may be due to the formation of
suboxides at the higher temperatures involved in welding.I4In any case, almost all commercial submerged arc
welding fluxes have very low FeO contents--typically
less than two percent.
North has claimed that CaO in the flux contributes
oxygen to the weld metal,' however, his fluxes were
made of blended minerals in which he used CaCO, as a
source of CaO. It is likely that the increasing oxygen
content with increasing CaO in North's fluxes is due to
carbonate decomposition and not CaO decomposition
as he concluded. In any case, little calcium is found in
the weld metal which confirms the fact that the
calcium reaction is relatively unimportant in welding.
Although free aluminum or free titanium in the
baseplate or welding electrode certainly reacts with the
oxygen in the weld metal, the quantity of these strongly
deoxidizing elements is generally low in carbon and low
alloy steels. Most welding experiments indicate that
little or none of these elements is recovered in the weld
metal, but that these elements are entirely consumed in
either the flux or as inclusions in the weld metal. Similar
arguments may be made for the other cations of the
typical flux components listed in Eq. [I].
The carbon-carbon monoxide reaction is important
as oxidation of carbon does increase the weld metal
METALLURGICAL TRANSACTIONS B

'

deoxidation rate, and hence influences the recovery of

Mn and Si; however, the work of Christensen3*shows
that the changes in Mn and Si are generally much
greater than the change in weld metal carbon content.
On this basis, the most important. but by no means the
only slag-metal reactions are those involving Mn and Si.
Using the above assumptions, it is possible to formulate a model of the thermodynamic equilibrium
which should prevail in flux shielded arc welding. An
outline of the calculation procedure follows.

- 28360
log K = -+ 10.61
T

K = - fisiO^

[21

asto,

where the standard state for SiO, in the slag is pure SiO,
and the standard state for oxygen and silicon in the
weld metal is based upon a 1 pet solution.

2. Calculation Procedure

Belton el a13have derived the following thermodynamic data for the Si02reaction in the temperature
range 17 13 to 2000 'C.

Table

I. Slag Systems and -Temperatures for

Thermodynamic Activity Data are Available

System

Temperature('C)

Which

Reference

Flux Basicity I n d e x
Fig. 3-The relationship between the equilibrium percent SiO, and
the flux basicity index for a given weld metal Si content in the
SO2-MnO-FeOsystem at 2000 'C.

0.5
1.0
F l u x B a s i c i t y Index

1.5

Fig. 2-The relationship between the activity of SiO, and the flux
basicity index for a given weld metal Si content at 2000 "Cassuming
regular solution behavior in the flux.
METALLURGICAL TRANSACTIONS B

F l u x Basicity I n d e x
Fig. &The relationship between equilibrium percent MnO and the
flux basicity index for a given weld metal Mn content in the
MnO-CaO-Al,03system at 2000 OC. The dotted lines are presented
for reference to discussion in the text.

VOLUME 12B,SEPTEMBER 1981-541

Using Assumption 1 and neglecting interaction terms

gives the relationship
Upon application of Assumption 2 and Eq. [3], the data
of Fig. 2 may be obtained. Using Assumption 3, and
available activity data for several flux systems given in
Table I, data such as Fig. 3 may be calculated.
A similar procedure may be used for the manganese
reaction beginning with

log K =

12760
- 5.68
T

In the first, slag-metal reaction data available from

the literature were used to determine if the theory is
capable of predicting the trend of Mn and Si transfer
for a wide range of flux, electrode and baseplate
compositions. This test is thought to demonstrate the
generality of the theory.
In the second, the accuracy of the theory was
investigated by meaurement of the extent of Mn and Si
transfer as a function of the electrode/baseplate composition. This test measures the ability of the theory to
predict the specific equilibrium compositions. As will be
shown subsequently, the theory withstood each of these
tests with remarkable success.

1. Verification of the Generality of the Theory

The manganese equation analogous to Eq.[3] is
Assumptions 2 and 3 again give a relationship for a
specific flux system of the type shown in Fig. 4.
Figures 3 and 4 represent the predicted equilibrium
between the Si or Mn in the weld metal and the SiO, or
MnO in the slag respectively, as a function of the
basicity index of the flux.*
* I n this paper, flux is used to describe the starting material prior
to welding while slag denotes the molten or fused flux during or
after welding.

The equilibria also vary with the flux system. A listing

of flux systems for which sufficient activity data are
available is given in Table I.
EXPERIMENTAL VERIFICATION
O F THE THEORY
Experimental verification of the proposed theoretical
equilibria is necessary if only because the assumptions
themselves are not easily verifiable. The value of any
theory is in proportion to its ability to provide accurate
predictions of the parameters of interest. T o this end
two separate verification techniques were used.

Consider the equilibria described by Fig. 3 and 4.

These apply specifically only to Si0,-MnO-FeO and
MnO-CaO-A1203fluxes; however, equilibria of a similar
nature may be generated for each of the flux systems
shown in Table I. It should be noted that the marked
dependence of equilibrium MnO or SiO, content as a
function of the flux BI is general, i.e., one always finds
the equilibrium flux composition to be strongly dependent upon the flux BI. As such, minor errors in
determining the flux BI should produce large errors in
the equilibrium prediciton.
As an example, consider a flux of composition 3 1 pet
CaO-4 pet MnO-65 pet A l a . The BI as given by Eq.
[ I ] is one. If a baseplate and electrode with average
initial composition 0.5 pet Mn is used for welding, the
equilibria of Fig. 4 would predict an equilibrium weld
metal content of 2 pet Mn. If the theory is correct one
would expect the weld metal to gain Mn by decomposition of part of the MnO in the slag during welding.
Fortunately, the slag BI does not differ appreciably
from that of the flux,18so that an iterative procedure is
not required. The final weld metal Mn content is
expected to lie between 0.5 pet and 2.0 pet. We may
define two quantitites, viz,
AMn = Mn, - Mn,

0.8

-1.2

-30

-20

-10

0
A MnO,%

10

20

T= 2000C

30

Fig. 5-Variation of change in weld metal manganese, AMn, with

deviation of flux MnO from predicted equilibrium, AMnO. Note that
all data points lie within the first and third quadrants.

542-VOLUME 12B, SEPTEMBER 1981

161

Fig. &Variation of change in weld metal silicon. ASi. with deviation

of flux SiO, from predicted equilibrium Asiai. Note that all data
points lie within the first and third quadrants.
METALLURGICAL TRANSACTIONS B

and
AMnO

[71

MnO, - MnO,,eq.

Where Mn, is the measured final weld metal manganese content, Mn, is the average initial electrodel
baseplate manganese content, MnO, is the initial flux
MnO content, and MnO,,^ is the percent MnO in the
slag which would be in equilibrium with the final weld
metal manganese as predicted by Fig. 4.
If in the example given Mn, = 1.0 pet, then AMn
= 1.0 - 0.5 = 0.5 pet and from Fig. 4 we find AMnO
= 4.0 - 1.8 = 2.2 pet. AMnO is a measure of the
deviation from equilibrium and may be considered as a
stimulus while AMn is the response to this stimulus. If
the initial flux contains an excess of MnO with respect
to the initial flux basicity and electrode/baseplate
equilibrium, MnO from the slag will decompose thereby
increasing the weld metal Mn content. The larger the
initial MnO deviation from the final predicted equilibriuim value, the greater is the stimulus for MnO
decomposition. As noted above, AMn is a measure of
the response to this stimulus. Reversing the order of the
initial-final subscripts between Eq. [6] and [7], ensures
that a positive stimulus will produce a positive response
and conversely a negative stimulus will provide a
negative response. In a physical sense, Eqs. [6] and [7]
are written such that a loss of manganese from the slag
as predicted by the theory (positive) will result in a gain
of manganese in the weld metal (also positive). If a
positive stimulus should produce a negative response, or
vice versa, the equilibrium predicition of the theory
would be in error.
Such tests of the theory may be made if one has the
following information:
a) The initial flux composition. Equation [l] is then

used to calculate a BI. To a very good approximation

the slag BI is assumed to be the same as the flux BI.I8
MnO, is determined from the flux composition.
b) The initial electrode and baseplate chemistries
and the final weld metal analysis and dilution. If the
dilution is not given, it may be assumed to be 0.5 for
most submerged arc welds.I9 Mn, and Mn, are taken
directly from these analyses. M n O , , is obtained from
an equilibrium plot such as Fig. 4.
The analysis described above has been performed for
over one hundred welds as reported in the literature.3-5-7'8J8-20"27
The data represent a wide range of
welding conditions with flux basicities varying from 0.2
to 4.0, and initial manganese and silicon contents
varying from 0.01 to 3.1 pet and 0.002 to 1.0 pet
respectively. Most welding fluxes were commercial
formultions containing seven or more components.
Since the available thermodynamic data is generally
only available for ternary or quarternary formulations
(Cf. Table I), only equilibria from flux systems comprising at least 80 pet of the commerical flux compositional anlysis were compared. This tends to exclude
welding fluxes with greater than 15 to 20 pet CaF,
although the analysis has been found to work reasonably well with these fluxes as well.28.
The results of the analysis of these welds are shown in
Fig. 5 and 6 for manganese and silicon respectively. It

I
0

0.5

1. 5

In1 tia l Manganese Content, percent

Fig. 8-The change in weld metal manganese content AMn, as a

function of the initial manganese content measured with a CaOMnO-SiO, flux using the artificial baseplate technique. The arrow
denotes the predicted equilibrium manganese content for this flux
system.

-"
mc

S 0.I
n

-.-VI"

c
.-

*0'
0

;-0.1
I

0.1

0.2

0.4
0.6
0.8
1.0
F l u x Basicity Index

1.2

Fig. 7-The effective weld pool temperature as measured by the SiO,

reaction equilibrium constant.
METALLURGICALTRANSACTIONS B

0.2
0.3
0.4
0.5
Initial Silicon C o n t e n t , percent

0.6

Fig. 9-The change in weld metal silicon content, ASi, as a function of

the initial silicon content measured with a CaO-MnO-SiO, flux using
the artificial baseplate technique. The arrow denotes the predicted
equilibrium silicon content for this flux system.
VOLUME 12B, SEPTEMBER 1981-543

Table Ii. Comparison of Theoretically Predicted Mn and Si Equilibrium with the Values Measured Experimentally

Flux Designation
Type
F- 1
CaO-MnO-SiO,

Flux Basicity
Index

Predicted
Mn, Pet

Measured
Mn, Pet

Predicted
Si Pct

Measured
Pct

Reference

1.17

1.83

1.85

0.33

0.28

This work
This work

F-4
CaO-SiO,

MIT-2
CaO-SiO,

0.46

1.1

1.27

1.45

0.52

0.2

0.15

0.15

0.52

<0.32

1.3

<0.5

0.80

0.95

This work

0.32

0.25

This work

0.25

This work

>1.18

This work

F-c
CaO-Si0,-Al,O,
F-d
CaO-SiOTAl,O,
F-e
CaO-MnO-SiO,
F-f
CaO-MnO-SiO,
n.d.--not determined.
All fluxes except F-1, F-2, MIT-1 and MIT-2 are commercial compositions.

Table 111. Comparison of Effective Reaction Temperature of Manganese Based Upon

Slag/Metal Surface Chemistries and Bulk Chemistries

Test
Number

Initial
Manganese
Mn,i, Pct

Final Surface
Manganese
Mn,.s, Pet

Final Bulk
Manganese
Mn,b, Pet

Final Bulk
Oxygen
0,b, pet

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

1.85
1.40
0.09
1.74
1.00
0.85
0.07
0.94
0.98
0.64
1.95
0.88
1.15
1.26
1.08
1.10
1.01
1.11
1.25
0.96

1.85
1.85
0.82
1.75
1.50
1.62
1.50
1.90
1.50
0.36
0.36
0.33
0.53
0.39
0.29
0.44
0.39

1.85
1.50
0.48
1.75
1.15
1.20
0.40
I .27
I .40
1.00
1.90
1.10
0.64
0.56
0.55
0.65
0.69
0.57
0.60
0.60

0.050
0.047
0.068
0.072
0.052
0.062
0.035
0.030
0.049
0.060
0.046
0.056
0.086
0.077
0.068
0.084
0.071
0.053
0.063
0.090

Surface
Equilibrium
Temperature
T.,'C

Bulk
Equilibrium
Temperature
T,,'C

1979
1968
1896
203 1
1950
1992
1885

1979
1932
1812
203 1
1905
1941
1694
1835
1928
1906
1968
1910
2164
2115
2088
2158
2138
2047
2088
2158

1968
1962
2050
203 1
1994
21 19
203 1
1929
203 1
2074

Tests 1 through 12 from this study.

Tests 13 through 20 from Christensen (Ref. 5)
Note the wider variation in reaction temperature when using bulk chemical analysis.

544-VOLUME

12B, SEPTEMBER 1981

METALLURGICAL TRANSACTIONS B

will be noted that none of the data falls within either the
II or IV quadrants. This confirms that the stimulus
predicted by the equilibrium theory agrees with the
response as determined experimentally. Considering the
imprecision with which one knows the actual weld
metal chemistry as well as the imprecision of the
thermodynamic data, the agreement of the theory and
experiment is quite remarkable.
It is worthy of note, that a similar analysis using
thermodynamic data based upon 1800 O C would not
yield good agreement, while an analysis based upon an
equilibrium temperature of 2200 'C, compresses the
data so as to obscure any trends. In addition, weld
metal equilibrium reaction temperatures based upon the
SiO-,reaction (Eq. [2]) agree with the 2000 OC assumption as may be seen in Fig. 7. Several of the welds of
Fig. 7 were preoduced with 3 pet silicon iron baseplate,
using very high basicity (low silica activity) fluxes. In
such a case one would expect transport of a large
amount of silicon from the weld pool to the slag. The
finding that the weld/slag equilibrium remains at
2000 OC suggests that silicon transport is very rapid
during submerged arc welding. The data confirm the
finding of Belton et aI3 that the SiO, reaction achieves
an effective equilibrium reaction temperature of
2000 OC over a wide range of weld metal chemistry and
flux composition.

2. Verification of the Accuracy of

the Equilibrium Mn and Si Predictions
Although the previous test confirms the generality of
the theory, it does not prove the accuracy of the
predictions. Indeed, there is considerable scatter of the
data within quadrants I and 111 of Figs. 5 and 6. This
scatter may be explained in part by the nonequilibrium
nature of slag-metal reactions in welding.29Nonetheless,
it is of interest to determine the accuracy with which the
theory can predict not only the gain or loss of Mn or Si
from the weld pool, but also the composition at which
change in composition is expected to be zero.

- 0.2

0.2

0.4

A S i , percent
Fig. 12-The predicted reltionship between ASi and ASiO, for
commercial flux L91. The prediciton indicates little possible change in
weld silicon content for initial values between 0.3 and 0.5 pet silicon.

0.5

1.5

I 1

I
2

2.5

n i t i o I Manganese Content, percent

Fig. 10Th change in the weld metal manganese content, AMn, as a

function of the initial manganese content measured with a CaOMnO-SiOTAl,03 flux using the artificial baseplate technique. The
arrow denotes the predicted equilibrium manganese content for this
flux system.

0.1

0.~2-~-0.3 0.4

0.5

0.6

0.7

Initial Silicon Content, percent

Fig. 11-The change in weld metal silicon content, ASi, as a function

of the initial silicon content measured with a CaO-MnO-Si0,-A120,
flux using the artificial baseplate technique. The arrow denotes the
predicted equilibrium silicon content for this flux system.
METALLURGICAL TRANSACTIONS B

- 0.2

0.2

Si

04

0.6

0.8

, percent

Fig. 13-The predicted relationship between ASi and ASiO, for

commercial flux L60. Most baseplates and electrodes contain less
than 0.5 pet silicon which would be increased by use of this flux.
VOLUME 12B, SEPTEMBER 1981-545

element constant. Although this ignores interaction

effects between the Mn and Si recoveries, it was found
to be a useful aproach. An example of two such series of
tests are given in Figs. 8 through 11. It is seen that the
Mn or the Si content in the weld deposit may vary
unpredictably when the electrode deviates significantly
from the equilibrium content but that the variation is
much more consistent near the equilibrium concentration. The arrows in Figs. 8 through 11 indicate the
predicted equilibria as obtained from graphs similar to
Figs. 3 and 4. Table I1 gives the measured and the
predicted equilibria for similar such tests made in our
laboratory and by others. It is seen that the agreement
is generally quite good, hence the accuracy of the
predicted equilibria is judged to be acceptable.
DISCUSSION

A Mn, percent
Fig. 14Th predicated relationship between AMn and AMnO for
commercial flux L60.

Several authors have attempted to study slag-metal

reactions during welding by producing a pyramid of
weld beads, of which each succeeding layer is diluted by
a smaller fraction of the base metal.l~5~8~39
In this way, it
is claimed that the final flux/electrode 'equilibrium'
may be attained. This is not necessarily true; the
pyramid of weld beads leads to a steady state weld
metal composition rather than an equilibrium composition. This is caused by the continuous disruption of
the weld pool chemistry by the melting electrode. Only
if the elecrode is of a composition which corresponds to
equilibrium with the flux will metal transfer between the
weld pool and the slag be zero. In all other cases, the
transfer reaches a steadv state which is a balance
between the rate of the slag-metal reaction and the rate
of weld pool dilution caused by the melting electrode.
In order to overcome the difficulties with the ovramid
bead test, it is desirable to have an electrode and
baseplate of identical composition; hence, dilution
effects may be neglected. Such identical electrode/
baseplate compositions are very difficult to obtain.
Instead, we have chosen to create an artificial baseplate
by bundling seven or eight lengths of electrode and
placing these on a water cooled copper hearth. in this
way, 'welds' may be made with an electrode and
baseplate of identical composition. A series of welds is
made, varying the Mn or the Si content of the electrode
until a given flux produces no net change in the deposit
chemistry as compared with the electrode chemistry.
This composition represents the composition of Mn or
Si in equilibrium with the flux. In practice it is
convenient to vary either the Mn or the Si content
without attempting to hold the composition of the other

546-VOLUME 12B, SEPTEMBER 1981

The above theory provides a means of calculating the

weld metal Mn or Si content which is in equilibrium
with a flux of a known composition. As such the theory
has been shown to be capable of predicting the direction, if not the magnitude, of the change in these
elements during welding.
Furthermore, the artificial baseplate experiments indicate that the approach to equilibrium is well behaved
for small deviations from the equilibrium but varies
considerably as one elects an electrode/baseplate composition which does not correspond closely to equilibrium with the flux. Recent results from our laboratory
indicate that the larger deviations from equilibrium
result in nonuniform distributions of Mn and sometimes of Si in the weld metal and of MnO and SiO, in
the slag. The distribution in the slag phase may be
markedly nonuniform; hence, the use of bulk composition values may lead to erroneous predictions in terms
of quantifying the extent of the Mn or the Si changes.
When one uses the surface compositions rather than the
bulk compositions of Mn and Si in the metal and slag
phases, the predictions correspond more closely to a
reaction temperature equilibrium of 2000 OC. Several
examples of the predicted equilibrium reaction temperatures based upon the bulk and the surface cornpositions of Mn are given in Table 111. It may be seen that
the temperatures based upon the surface compositions
are more uniform and are centered about an effective
reaction temperature of 2000 O C . This provides a furtherjustification for assumption l.
As presented the analysis applies only to fused,
neutral submerged arc welding fluxes with CaF, content
of less than 20 pet* although several tests indicate that
*The restriction to systems with less than 20 pet CaF, is based
upon the limited activity data for CaF2-metal oxide slags.

the approach may be applicable to fluxes with higher

CaF2 contents as well. Comparison of the predictions
with the results of other flux shielded welding processes
would be welcomed. Extension of the experimental
work to blended submerged arc fluxes is underway in
our laboratory. The preliminary results are encouraging
that the techique may be useful for these systems as
well. If successful, this work, coupled with a mass
balance formalism, will hopefully provide a means of
METALLURGICAL TRANSACTIONS 8

predicting Mn and Si transfer in active submerged arc

fluxes.
Although the study as presented is only approximate,
and has been verified using aposteriori data, it may be
of use in selecting combinations of flux/electrode/baseplate apriori. Consider for instance, the predicted
equilibria of Fig.12 through 14, which represent two
commercial fluxes. The predicted equilibrium silicon
content for flux L91 is approximately 0.4 pet Si.
Electrode/baseplate combinations below this value
should gain silicon in the weld metal when this flux is
used, while values greater than 0.4 pet would lead to
loss of silicon during welding. The initial MnO in flux
L91 is essentially zero suggesting that this flux would
always lead to loss of manganese from the weld pool.
Flux L60 would be relatively inert when using an
electrode/baseplate of average composition 1.27 pet Mn
and 0.85 pet Si. While these graphs do not predict the
exact composition of the weld metal to be deposited,
they may be used to predict the permissable variations.
Although in many cases these variations are quite wide,
in other cases, the predicted range restricts the possible
weld metal compositions considerably. It is anticipated
that these more restrictive flux/electrode/baseplate
compositions may perform more reproducibly in practice.
SUMMARY
A theory predicting slag-metal equilibrium during
submerged arc welding with fused neutral fluxes has
been presented. Although not capable of predicting the
magnitude of the Mn and Si changes in the weld metal,
the theory is capable of predicting the gain or loss of
these elements over a wide range of flux/electrode/
baseplate compositions. As such, it represents a significant advance in our ability to predict and control the
extent of these reactions. Further studies relating this
work to other types of submerged arc welding fluxes
and other welding processes is desirable.
ACKNOWLEDGMENTS
The authors gratefully acknowledge support for this
study from the National Science Foundation under
grant DMR77-27230. They are also grateful to Dr.
Harry Bishop of Allegheny-Ludlum Steel Corporation
for supplying the three percent silicon iron plate used
for several of the trials.
REFERENCES
1. N. Christensen and J. Chipman: Weld. Res. Counc. Bull.. January
1953, no. 15.
2. R. A. Kubli and W. B. Sharav: Weld.J., 1961, vol. 40, no. 11, p.
4974.

METALLURGICAL TRANSACTIONS B

3. G. R. Belton, T. J. Moore, and E. S. Tankins: Weld. J., 1963,

vol. 42. no. 7, D. 289-s.
4. W. J. Lewis and P. J. Rieppel: Weld.J.. 1961, vol. 40, no. 8, p.
337-s.
5. N. christensen: Report AD-602138, NTIS, Arlington. VA, November. 1965.
6. C. A. Butler and C. E. Jackson: Weld.J., 1967, vol. 46, p. 448-s.
7. T, H. North, H. B. Bell, A. Nowicki, and I. Craig: Weld.J., 1978,
vol. 57, no. 3, p. 63-s.
8. T. H. North: Weld. Res. Abroad, January 1977, vol. 23, no. 1, p. 2.
9. C. E. Jackson: Weld. Res.Counc., December 1973, no. 190.
10. J. G. Garland and N. Bailey: M/84/75 The Welding Institute,
Abington, England, 1975.
11. J. H. Palm: Weld.J., 1972, vol. 5 1, no. 7, p. 358-s.
12. T.Boniszewski: Met. Const. Br. Weld.J., 1974, vol. 6, p. 128.
13. T, W. Eagar: Weld.I., 1978, vol. 57, no. 3, p. 76-s.
14. T.W. Eagar: Weldments: Physical Metallurgy and Failure Phenomena. R. J. Christofel, E. F. Nippes, and H. D. Solomon, eds.,
p. 31, General Electric Co., Schenectady, NY, 1979.
15. B. A. Korh: Autom. Weld., 1977, vol. 30, no. 7, p. 16.
16. J . D. Cobine and E. D. Burger: J. Appl. Phys., 1955, vol. 26, no. 7,
p. 895.
17. 0 . H. Nestor: J . Appl. Phys., 1962, vol. 33, no. 5, p. 1638.
18. S. S. Tuliani, T. Boniszewski, and N. F. Eaton: Weld. Met. Fabr.,
1969, vol. 37, no 8, p. 327.
19. B. G. Renwick and B. M. Patchett: Weld.I., 1976, vol. 55, no. 3.
p. 69-s.
20. J. G. Garland and P. R. Kirkwood: Welding of Line Pipe Steels,
K. H. Koopman, ed., p. 176, Welding Research Council, New
York, 1977.
21. S. F. Baumann, J. R. Sawhill, and M. Nakabayashi: ibid, p. 56.
22. W. K. C. Jones: Weld.J., 1976, vol. 55, no. 2, p. 42-s.
23. A. P. Bennett and P. J. Stanley: Br. Weld. J., 1966, vol. 13, no. 2,
p. 59.
24. G. Uttrachi: private communication, Union Carbide Corp.,
Astabula, OH, 1977.
25. J. G. Garland and P. R.Kirkwood: Weld. Met. Fabr., 1976, vol.
44, no. 4, p. 217.
26. N. Bailey: Weld. Res. Int., 1978, vol. 8, no. 3, p. 215.
27. J. G. Garland and N. Bailey: ibid, 1978, p. 240.
28. C. S. Chai and T.W. Eagar: unpublished research, MIT, Cambridge, MA, 1977.
29. C. S. Chai and T, W. Eagar: Weld.J., 1980, vol. 59, no. 3, p. 93-s.
30. J. F. Elliott, M. Gleiser, and V. Ramakrisna: Thermochemistryfor
Steelmaking, Vol. 11, Addison Wesley Publishing Co., Reading,
MA, 1963.
3 1. F. D. Richardson: Physical Chemistry of Melts in Metallurgy, Vol.
I, Academic Press, NY 1974.
32. N. Christensen: Jernkontorets Ann., 1977, vol. JkA-77, no. 5, p. 4.
33. H. Fujita and S. Maruhaski: Tetsu To Hagane, 1970, vol. 56, p.
830.
34. S. Mamhashi: Teisu To Hagane, 1971, vol. 57, p. 891.
35. R. H. Rein and J. Chipman: Trans. TMS-AIME, 1965, vol. 233, p.
415.
36. E. Martin, 0.1. H. Abdelkarim, I. D. Somerville. and H. B. Bell:
Metal-Slag-Gas Reactions and Processes, Z . A. Foroulis and W. W.
Smeltzer, eds., p. 1, The Electrochemical Society, Princeton, NJ,
1975.
37. The Making, Shaping and Treating of Steel, H . E. McGannon, ed.,
9th ed., U.S. Steel. 1971.
38. J. G. Garland and P. R. Kirkwood: Met. Constr., 1975, vol. 7 , no.
6, p. 320.
39. H. Thier: Proceedings of the Conference on Weld Pool Chemistry
and Metallurgy, p. 271, The Welding Institute, London, April
1980.

VOLUME 12B, SEPTEMBER 1981-547